Running model.eval() function

ssalome · February 27, 2020, 8:47am

Hi
I have this doubt
I want to create a confusion matrix and I’m a beginner. I’m getting as error at
This is evaluation function

Blockquote
pred = torch.tensor()
confusion_matrix = torch.zeros(num_classes, num_classes)
def evaluation(dataloader, model):
total, correct = 0,0
for data in dataloader:
inputs, labels = data
inputs, labels = inputs.to(device), labels.to(device)
# inputs are put in the gpu
outputs = model(inputs)
_, pred = torch.max(outputs.data,1)
for t, p in zip(classes.view(-1), preds.view(-1)):
confusion_matrix[t.long(), p.long()] +=1
total += labels.size(0)
correct +=(pred==labels).sum().item()
return 100 * correct/total

This is the training loop

Blockquote
loss_epoch_arr =
max_epochs = 1
min_loss = 1000

n_iters = np.ceil(1050/batch_size)

for epoch in range(max_epochs):
for i, data in enumerate(trainloader, 0):

    inputs, labels = data
    inputs, labels = inputs.to(device), labels.to(device)

    #print(inputs.shape)
    opt.zero_grad()

    outputs = resnet(inputs)
    loss = loss_fn(outputs, labels)
    loss.backward()
    opt.step()

    if min_loss> loss.item():
        min_loss = loss.item()
        best_model = copy.deepcopy(resnet.state_dict())
        print('Min loss %0.2f' % min_loss)

    if i%100 == 0:
        print('Iteration : %d/%d, Loss : %0.2f' % (i, n_iters, loss.item()))

    del inputs, labels, outputs
    torch.cuda.empty_cache()

loss_epoch_arr.append(loss.item())

print('Epoch: %d/%d, Test acc: %0.2f, Train acc : %0.2f' % (epoch, max_epochs, evaluation(testloader,resnet), 
                                                            evaluation(trainloader, resnet)))

plt.plot(loss_epoch_arr)
plt.show()

I’m getting the following error

Blockquote

AttributeError Traceback (most recent call last)
in
32 loss_epoch_arr.append(loss.item())
33
—> 34 print(‘Epoch: %d/%d, Test acc: %0.2f, Train acc : %0.2f’ % (epoch, max_epochs, evaluation(testloader,resnet),
35 evaluation(trainloader, resnet)))
36 plt.plot(loss_epoch_arr)

in evaluation(dataloader, model)
11 # from net to model - 2 change from lenet
12 _, pred = torch.max(outputs.data,1)
—> 13 for t, p in zip(classes.view(-1), preds.view(-1)):
14 confusion_matrix[t.long(), p.long()] +=1
15 total += labels.size(0)

AttributeError: ‘tuple’ object has no attribute ‘view’

Blockquote

Can someone help me with this?

phan_phan · February 27, 2020, 9:49am

In the line

for t, p in zip(classes.view(-1), preds.view(-1)):

either classes or preds (or both) is a tuple, when it should be a tensor.

It might be preds, that should maybe be pred, because on the line just before we have _, pred = torch.max(outputs.data,1)

Or it might be classes, that should maybe be labels, because a few lines before we get the tensor labels that is never used.

Or maybe both ?

ssalome · February 29, 2020, 4:44am

yeah u were right it is supposed to be labels, labels is tensor, preds was already tensor.
But i have a doubt i rougly ran it for 1 epoch to see if the code for confusion matrix works…I had defined it in evaluation function…like on loading the model and then running the evaluation function on
train and test loader it should give some non- zero values but I’m getting all zeros only. What could be the error ?

ssalome · February 29, 2020, 4:47am

This is the evaluation function

phan_phan · February 29, 2020, 12:32pm

preds, defined at the first line, is an empty tensor
pred, defined in your function, is the one containing the predictions
To compute the confusion matrix, only preds is used.
I think the confusion lies there

ssalome · March 1, 2020, 12:17pm

Yeah the confusion was probably that only. Thank you:)

ssalome · March 5, 2020, 9:11pm

hi
Now Im able to get the probabilities but there’ s a problem i have only 1493 images in total and 1045 go for training and 448 for testing…but the confusion matrix rows sum up to 5100…what can be the error.

#preds = torch.tensor()
confusion_matrix = torch.zeros(num_classes, num_classes)
def evaluation(dataloader, model):

model added as argument - 1 change from lenet

total, correct = 0,0
for data in dataloader:
    inputs, labels, imgindex  = data
    inputs, labels, imgindex = inputs.to(device), labels.to(device), imgindex
    # inputs are put in the gpu
    outputs = model(inputs)
    # from net to model - 2 change from lenet
    _, pred = torch.max(outputs.data,1)
    for t, p in zip(labels.view(-1), pred.view(-1)):
        confusion_matrix[t.long(), p.long()] +=1
    total += labels.size(0)
    correct +=(pred==labels).sum().item()
return 100 * correct/total